linux-kernel - Re: [PATCH] perf script python: Fix dict reference counting

lists.openwall.net		lists / announce owl-users owl-dev john-users john-dev passwdqc-users yescrypt popa3d-users / oss-security kernel-hardening musl sabotage tlsify passwords / crypt-dev xvendor / Bugtraq Full-Disclosure linux-kernel linux-netdev linux-ext4 linux-hardening linux-cve-announce PHC
Open Source and information security mailing list archives

Hash Suite: Windows password security audit tool. GUI, reports in PDF.

[<prev] [next>] [<thread-prev] [thread-next>] [day] [month] [year] [list]

Date:   Mon, 9 Jul 2018 12:25:33 +0300
From:   Janne Huttunen <janne.huttunen@...ia.com>
To:     Jiri Olsa <jolsa@...hat.com>
CC:     <linux-kernel@...r.kernel.org>,
        Alexander Shishkin <alexander.shishkin@...ux.intel.com>,
        Andi Kleen <ak@...ux.intel.com>,
        Arnaldo Carvalho de Melo <acme@...nel.org>,
        Jaroslav Škarvada <jskarvad@...hat.com>,
        Namhyung Kim <namhyung@...nel.org>,
        Peter Zijlstra <peterz@...radead.org>
Subject: Re: [PATCH] perf script python: Fix dict reference counting

On Sun, 2018-07-08 at 13:17 +0200, Jiri Olsa wrote:
> On Fri, Jul 06, 2018 at 09:53:44AM +0300, Janne Huttunen wrote:
> > 
> > The dictionaries are attached to the parameter tuple that steals the
> > references. The code should not decrement the reference counters
> > explicitly. Otherwise the objects might be released while they are
> > still in use which may cause perf crashes, assertions or just plain
> > weird behavior like unexpected data changes in stored objects.
> > 
> > Signed-off-by: Janne Huttunen <janne.huttunen@...ia.com>
> > ---
> >  tools/perf/util/scripting-engines/trace-event-python.c | 8 ++------
> >  1 file changed, 2 insertions(+), 6 deletions(-)
> > 
> > diff --git a/tools/perf/util/scripting-engines/trace-event-python.c b/tools/perf/util/scripting-engines/trace-event-python.c
> > index 46e9e19..60fce44 100644
> > --- a/tools/perf/util/scripting-engines/trace-event-python.c
> > +++ b/tools/perf/util/scripting-engines/trace-event-python.c
> > @@ -908,14 +908,11 @@ static void python_process_tracepoint(struct perf_sample *sample,
> >  	if (_PyTuple_Resize(&t, n) == -1)
> >  		Py_FatalError("error resizing Python tuple");
> >  
> > -	if (!dict) {
> > +	if (!dict)
> >  		call_object(handler, t, handler_name);
> > -	} else {
> > +	else
> >  		call_object(handler, t, default_handler_name);
> > -		Py_DECREF(dict);
> > -	}
> >  
> > -	Py_XDECREF(all_entries_dict);
> >  	Py_DECREF(t);
> >  }
> >  
> > @@ -1235,7 +1232,6 @@ static void python_process_general_event(struct perf_sample *sample,
> >  
> >  	call_object(handler, t, handler_name);
> >  
> > -	Py_DECREF(dict);
> >  	Py_DECREF(t);
>
> so the dict is released when the tuple is released?

To the best of my knowledge, yes.

As far as I can see, there is only a single reference to each dict
and according to the Python documentation PyTuple_SetItem() "steals"
the reference passed to it. If so, afterwards the tuple owns the
only reference to the dict(s) and should take care of releasing
them when appropriate.

I even built libpython with reference debugging enabled and when I
run perf without the fix I get this:

Fatal Python error: Objects/tupleobject.c:238 object at 0x7f10f2041b40 has negative ref count -1
Aborted (core dumped)

With the fix I get no errors.